Robert-Jan Bruintjes

PhD candidate with Jan van Gemert @ Computer Vision Lab, TU Delft

I study Deep Neural Networks for Computer Vision, like CNNs and Vision Transformers, and how they learn to do their job. Topics of choice include learned equivariance, position information and the inductive bias of the self-attention operation. I’ve also dabbled in implicit neural convolutional operators a bit.

Highlights

Recent

Report: VIPriors challenges 2023

June 2024 We published a report on the challenge track of VIPriors 2023.

VISAPP 2024 paper: Scale Learning

February 2024 See publication page.

NeurIPS 2023 paper: Color Equivariant Convolutions

September 2023 Co-authored paper with colleagues at TU Delft.

More updates

Featured Publications

Publication January 2022 ICLR 2022
FlexConv: Continuous Kernel Convolutions with Differentiable Kernel Sizes

FlexConv: Continuous Kernel Convolutions with Differentiable Kernel Sizes

We provide a high bandwidth, alias-free convolutional kernel parameterization with learnable kernel size and constant parameter cost.
ArXiv Code Poster Slides
Publication April 2023 CVPRW 2023
What Affects Learned Equivariance in Deep Image Recognition Models?

What Affects Learned Equivariance in Deep Image Recognition Models?

We find evidence that learned equivariance in image recognition models correlates with accuracy, then we show how to increase said equivariance.
ArXiv Poster

Publications

VIPriors 4: Visual Inductive Priors for Data-Efficient Deep Learning Challenges
ArXiv Robert-Jan Bruintjes, Attila Lengyel, Marcos Baptista Rios, Osman Semih Kayhan, Davide Zambrano, Nergis Tomen, Jan van Gemert
Website ArXiv Code
Scale Learning in Scale-Equivariant Convolutional Networks
VISAPP 2024 Mark Basting, Robert-Jan Bruintjes, Thaddäus Wiedemer, Matthias Kümmerer, Matthias Bethge, Jan van Gemert
PDF DOI
Color Equivariant Convolutional Networks
NeurIPS 2023 Attila Lengyel, Ombretta Strafforello, Robert-Jan Bruintjes, Alexander Gielisse, Jan van Gemert
ArXiv Code
Benchmarking Data Efficiency and Computational Efficiency of Temporal Action Localization Models
ICCVW 2023 Jan Warchocki, Teodor Oprescu, Yunhan Wang, Alexandru Damacus, Paul Misterka, Robert-Jan Bruintjes, Attila Lengyel, Ombretta Strafforello, Jan van Gemert
ArXiv DOI
VIPriors 3: Visual Inductive Priors for Data-Efficient Deep Learning Challenges
ArXiv Robert-Jan Bruintjes, Attila Lengyel, Marcos Baptista Rios, Osman Semih Kayhan, Davide Zambrano, Nergis Tomen, Jan van Gemert
Website ArXiv Code DOI
See all publications

Contact

I am always willing to talk about collaboration or supervision. Please find my contact info at the top of the page.

Office

TU Delft, building 28, room 6.E.280
Van Mourik Broekmanweg 6
2628 XE Delft
The Netherlands